Targeted advertising has recently grown such that advertisements are provided based on a user's online search sessions, images viewed by the user, contents of e-mails, etc. Advertisements that are targeted to users based on video content, however, have traditionally been selected based on a type of web page where the video content appears (e.g., news page, entertainment page), based on the genre to which the video content belongs, or even based on a time marker in a video stream in which an advertiser believes an advertisement would be relevant. However, a time marker is ineffective when a video is montaged, divided up into various segments, etc. While object recognition within video content has also been used to target advertisements, object recognition is computationally heavy and time intensive.